Experiences with fine-grained parallel genetic algorithms

In this paper, we present some results of our systematic studies of fine-gr ained parallel versions of the island model of genetic algorithms and of va riants of the neighborhood model (also called diffusion model) on the massi vely parallel computer MasPar MP1 with 16k processing elements. These paral lel genetic algorithms have been applied to a range of different problems ( e.g. traveling salesman, capacitated lot sizing, resource-constrained proje ct scheduling, flow shop, and warehouse location problems) in order to obta in an empirical basis for statements on their optimization quality.